Validity of modified methods to assess three welfare indices for use in on-farm pig welfare monitoring

The Australian Pork Industry Quality Assurance Program (APIQ✓^®) requires producers to be annually audited against a set of standards and performance indicators, but does not provide them with the opportunity to monitor and (or) benchmark the welfare status of their animals over time. Although the assessment of animal welfare at farm level remains an on-going challenge, the literature demonstrates the opportunity to develop a practical welfare assessment tool, using valid and reliable animal-based welfare indices (Winckler et al. 2003; Knierim and Winckler 2009). Whilst extensively employed within the literature and more recently within on-farm assessment schemes, the validity of animal-based welfare indices modified from experimental settings for practical use in on-farm welfare monitoring has not been robustly examined. This study aimed to examine, in a commercial setting, the validity and intra- and inter-observer reliability of modified methodologies (M) of three commonly used animal-based pig welfare indices: body condition score (BCS), lameness score (LS) and injury score (IS).

To improve on-farm practicality and reduce observer subjectivity, validated methods of assessment (V) for the three welfare indices were simplified to create the modified methodologies (BCS: Patience and Thacker 1989; LS: Karlen et al. 2007; IS: De Koning 1985). For example, BCS was modified from a 5-point visual and tactile assessment of the pig’s condition performed outside of the group-pen, to a 3-point visual assessment of the pig conducted within the group-pen. The validity and reliability of the M measures of BCS, LS and IS were investigated in group-housed sows and grower pigs at a large Australian commercial piggery, over a 6-week period. Four trained observers sampled 240 group-housed pigs over six 2-day periods (120 sows in weeks 1–3 and 120 grower pigs in weeks 4–6); each observer assessed 20 focal animals for BCS, LS and IS on d 1 using both the M and V methodologies, and 40 focal animals using the M methodologies on d 2. Whilst not blind to the group on d 2, observers were blind to the individual animal.

The validity of the M methodologies was investigated using Spearman’s rho correlations (ρ) to examine the strength of the relationship between the assessment scores from the M and V measures, and Kappa statistics (κ) to determine the level of agreement between the two measures. The reliability of the M methodologies was investigated using a test-retest assessment that used ρ to examine the similarity between measures collected on an animal at two different time points (intra-observer), and κ to investigate the agreement between measures taken on an animal by multiple assessors (inter-observer).

Moderate to substantial levels of agreement (κ = 0.61 to 1.00) confirmed the intra- and inter-observer reliability of the M methodologies in group-housed sows and grower pigs. However, validity testing only indicated a moderate relationship (ρ = 0.30 to 0.49) with slight to fair levels of agreement (κ = 0.21 to 0.60) between the M and V methodologies. Given that the V indices underwent only minor modification, greater correlation and agreement between the measures were expected. These results may be due to the homogeneity of the data due to a lack of variation in the condition of the animals sampled, rather than a genuine lack of validity of the M methodologies. The lack of variation in the sample means that the minor inconsistencies that are commonly found between observers/observations are enough to substantially reduce the level of agreement between the measures. Whilst the correlations and level of agreement between the measures were not as strong as expected, the current findings do not refute the validity of the M methodologies as on-farm measures of BCS, LS and IS in group-housed sows and grower pigs.

Given these findings, further testing in populations with greater variation is required to confirm the validity of the M methodologies. Confirming the validity of these measures is vital if they are to be used effectively by producers to monitor and benchmark pig welfare over time.